Contourlet变换域的稀疏表示分类方法  

Sparse Representation for Classification Based on Contourlet Transform

作　　者：廖传柱[1] 潘婷婷[2] 江铭炎[3] 

机构地区：[1]漳州职业技术学院电子工程系,福建漳州363000 [2]中国移动通信集团江苏有限公司,江苏南京210029 [3]山东大学信息科学与工程学院,山东济南250100

出　　处：《内蒙古师范大学学报（自然科学汉文版）》2016年第1期89-93,共5页Journal of Inner Mongolia Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61201370);山东省自然科学基金资助项目(ZR2014FM039)

摘　　要：提出一种基于Contourlet变换域的稀疏表示分类方法对人脸进行识别,使用Contourlet波变换对初始图像进行处理,得到原始图像的低频和高频特征,将低频分量与高频分量直接组合为一维向量,输入稀疏表示分类算法进行识别.研究结果表明,该方法能够对图像进行快速特征提取,去除噪声和冗余,保留边缘等局部特征,同时降低了图像维数.与PCA+SRC、LPP+SRC方法相比,该方法能够得到更好的判别特征和更高的识别率.A kind method based on sparse representation and the Contourlet transform for face recognition is proposed.The original images are filtered by the Contourlet transform,the low and high frequency characteristics are obtained.Then combining these two kinds of characteristics into one dimensional vector,the vector is put into the sparse coding algorithm for the face recognition.The results shows that the proposed method can process the image feature extraction rapidly,remove noise and redundancy,retain local image features,and reduce the data dimensionality.The proposed method can obtain better features and better face recognition performance by comparison with PAC＋SRC method and LPP＋SRC method.

关 键 词：稀疏表示分类方法 CONTOURLET变换 低频子带 高频子带 人脸识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]